Rheumatoid Arthritis (RA) is a prevalent disease (1% of the population), characterized by chronic inflammation and destruction of joints, predominantly in the hands and feet. Current therapeutics can significantly reduce joint damage (~50%) and chronicity (~20%), thereby increasing quality of life and preventing high healthcare costs. This is only possible if treatment is initiated in a very early stage. Since recent studies suggest that a dedicated extremity-MRI scanner could detect inflammation before it is clinically detectable, it may become an essential tool for achieving these very early diagnoses. However, the interpretation of MRI is hampered by a lack of efficient, sensitive and specific imaging biomarkers, thereby not making full use of these powerful treatments. Reasons for this are that a) MRIs are evaluated by a visual scoring method that requires three months of training, whilst the analysis itself is also time-consuming (30 min./patient); b) image quality is limited by image inhomogeneities and because the small-bore permanent magnet requires scanning in different stations; c) sensitivity and reproducibility are inadequate because the expected subtle changes are difficult to detect by the human visual system; d) specificity is restricted since healthy subjects can also have subtle abnormalities. Goals The goal of this project is to develop and validate a computer-aided system for very early detection of RA from extremity MRI. To realize this, we will address the following major technological challenges: a) We will investigate computer-assisted time-efficient analysis algorithms for user-friendly visualization of the MRI data, comparing to image data from healthy individuals. b) We aim to improve image quality by inhomogeneity correction and image stitching, and by developing methods for stabilizing the hand and foot during image acquisition. c) To increase sensitivity and reproducibility, we aim to automatically quantify imaging biomarkers (such as bone marrow edema, erosions, synovitis) that are known to be visible in more advanced RA, but may already be detectable in an earlier stage by intensity- and texture-based analysis. d) To improve sensitivity and specificity, we aim to construct a reference image atlas to discover more subtle changes, while accounting for normal anatomical variations. This will require an articulated atlas-based approach with affine and elastic image registration. We will follow two concurrent strategies; extracting features from the intensity differences between patient and corresponding normal reference image data and subsequently applying pattern recognition; and by adaptive, semi-supervised classifiers. e) Finally, the developed methodology will be validated extensively in longitudinal studies, to determine the predictive value. While effective therapeutics for RA have become available, a very early diagnosis is essential for preventing joint damage and chronicity. Recent clinical research suggests that MR imaging can contribute significantly to a very early diagnosis, but this requires technological developments in the automated interpretation of these large image data. Addressing these technical issues will have a profound impact on patient?s quality of life and healthcare costs. As a result, several stakeholders have interest in the results of this project: " Patients: Because of the envisioned increase in quality of life of RA patients, preventing the development of RA into a chronic disease, the Reumafonds is interested in implementing the results into clinical practice and has committed to a €50k in-cash contribution to this project. " Clinical end-users: Clinical partners from Rheumatology and Radiology, who have experience in using prototype software, will evaluate the methodology and assure the clinical relevance of the developments. " Medical device industry: Upon successful completion of this project, the software from this project may become available commercially for clinical use, through Medis medical imaging systems BV, after further software development, quality control, FDA clearance etc. Therefore, Medis will contribute €30k in-cash and €15k in-kind to this project. PS-Tech is interested in the project results as they are specialized in presenting complex 3D datasets (such as extremity-MRI data) through virtual reality workstations with intuitive user interaction tools. As such PS-Tech has committed to a €29,750 in-kind contribution. " Clinical research organizations (CRO): Since the developed comparative imaging techniques can be used to monitor the progression of early RA, the quantification can also be used for drug efficacy trials by the pharmaceutical industry. As a CRO, BioClinica could offer their services to pharmaceutical companies using the developed software, and therefore committed to a €15k in-kind contribution. Percuros BV is operating in the area of drug discovery research, p.a. in skeletal biology, and will contribute €12.5k in-cash. Using the expertise of the user group, utilization of the developed techniques in early diagnosis of other disease like osteoarthritis or spondyloarthritis will be explored after finalizing the project.
